You are viewing Singapore’s website
The dimension missing from contact center data. A way to classify customer interactions that answers the question most taxonomies skip: why now.
Two Chinese sayings contradict each other. A contact center proves both of them right.
The first: without wind, there are no waves (无风不起浪). A wave is evidence of something that pushed it. The second runs the other way: even with no wind, the sea swells three feet (无风三尺浪). Some of the motion answers to nothing upstream. The water moves on its own.
I spent months on how a contact center produces its volume, and the two sayings kept mapping onto two halves of the same operation. That is where this classification came from. It is built around a question the standard schemes leave out.
What current taxonomies leave out
Standard classification answers two things well. How many interactions, by category. How they arrived, by channel and queue. For an operation graded on throughput, that covers it. For an agency that runs the upstream marketing and CRM and the downstream service, it drops the question that decides the economics: why did this interaction happen now.
A concrete case. A brand changes a loyalty rule so points expire at year end. For weeks afterward, customers contact the brand to redeem before the value disappears. Standard tagging records "points redemption" as the topic and "inbound voice" or "Tmall chat" as the channel. Accurate labels. Neither one captures the fact that the whole surge came out of a single CRM decision. Design and run that loyalty program and you caused the surge. Staff only the queues and you are catching a wave that someone upstream set off, with nothing in your data to tell you so.
That gap costs money. Marketing and service look at one surge and file two unrelated reports. Marketing books the redemptions as a campaign result. Service books the same volume as a headcount shortfall. Nobody connects them, because the data was never built to connect them.
Three axes, kept apart
The repair is to stop asking one tag to carry three meanings. Each interaction is classified on three separate dimensions.
Wind: what caused it. The upstream trigger, if there is one.
Wave: what it concerns. The topic, the ordinary call driver that operations teams already track.
Channel: how and where it reaches you. WeCom, Tmall, a hotline, a chatbot.
They stay independent. One wind raises several different waves: the loyalty change above produces redemption requests, fairness complaints, and a steady trickle of customers surprised they held any points at all. One wave shows up with or without wind: a sizing question can follow a livestream or land on a slow Tuesday for no traceable reason.
Channel is the one people miscast as a cause, because the correlation is right there. Open a Tmall store, watch contact volume climb, conclude that Tmall drives the volume. Tmall is the shoreline the waves break on. The wind is elsewhere. A shopper in that store could be there because of an hour-old livestream or because she wants to know whether a coat runs small. One channel, two unrelated causes. Read the channel as the trigger and the causal question disappears. This is the most common error in contact center taxonomies: marketplace, brand store, and retail get listed as if they were reasons for contact, when they are places contact lands.
The triggers, sorted by whether there is wind
Strip the channel back out and the causes sort into two bands.
The top band is event-driven. Something specific happened upstream, and a surge followed. The bottom band is structural, the standing conditions of being a brand that sells things and keeps customers. This is the three-foot swell that arrives with no wind at all.
The interaction trigger map: triggers by engagement mode (Reactive, Proactive, Continuous), split into event-driven (有风) and structural baseline (无风) bands.
The three engagement modes across the top are where the integrated practice separates from everyone else. Reactive is consumer-initiated service, the column every contact center already staffs. Proactive is brand-initiated: the outbound call, the expiry reminder, the win-back. Continuous is the ongoing relationship: clienteling, VIP concierge, private domain nurture that has no start and no ticket.
Traditional contact center providers live almost entirely in the reactive column. eCommerce service shops cover reactive plus some proactive, mostly inside marketplace transactions. A practice that runs the upstream programs and the downstream service is the only kind that operates across every trigger and all three modes. The matrix is how you show that on one page, with a brand's own volume in the cells.
Why the second saying earns its place
Look again at the bottom band. It is tempting to treat the no-wind rows as leftover noise and chase only the event-driven spikes. That gets the value backward.
Baseline is the reference line that makes every wind measurable. Normal week runs a thousand contacts; a campaign week hits eighteen hundred; the campaign owns the eight hundred extra. That difference is the honest figure for what marketing handed to service. Without a clean baseline, you cannot say what any single wind produced. The flat weeks are what let you price the busy ones.
What it lets you ask
With all three axes in place, questions that used to be guesswork turn into queries. What share of total volume comes from programs we run, against the baseline we would carry regardless. Which campaigns drop the heaviest service load, and whether the campaign budget ever accounted for it. What customers raise in the days after a CRM message goes out. Where self-service pays off, which tends to be the high-volume, no-wind, routine topics.
The first question is the one that moves the commercial conversation. Once a brand can see what share of its own contact volume its upstream programs generate, the build-versus-buy and the bundling question both change shape. The marketing operation and the service operation stop looking like two purchases to split apart.
A model reads each interaction after it closes and applies the tags. Agents do not pick from a dropdown mid-call; they are in no position to judge cause while the conversation is live, and dropdown fatigue rots the data inside a month regardless. The channel needs no model at all. The platform already records it.
The test is plain. Take ninety days of a brand's own interactions, tag them on these three axes, and read off the share that traces to programs the brand already pays for. We run that as a workshop, against real volume, and the cells fill in fast.
*Daniel Wei is COO, TP Infinity, the consumer engagement practice within TP. The Wind / Wave / Channel framework is part of the practice's interaction analytics work across Asia Pacific and beyond.
